Aurum Risk Core Gold Protection Engine

Your EA does not lose money because the entry was wrong. It loses because nothing stopped it. A bad signal costs you one trade. A missing risk layer costs you the account. Yet almost every gold EA spends its complexity on entries and treats protection as an afterthought - a stop-loss input and a comment about money management. Aurum Risk Core is the other half. It is the chassis: sizing, stops, basket control, exposure limits, entry gating, restart safety, and the instrumentation to see all of it working. You supply the entry. This is a framework, not a signal. It makes no performance claim, and the section below explains exactly why. --- Why gold specifically --- XAUUSD punishes what other symbols forgive. Spreads that jump from 15 to 80 points in a second. Stop levels that move under you. Volatility that makes a fixed lot size reckless at 09:30 and pointless at 03:00. A risk layer written for EURUSD will not survive here. This one was built, tested and broken against gold tick data exclusively. --- 1. Limits that cannot be raised --- Every protective ceiling is a compile-time constant. Setting an input above it does not work: the value clamps and the change is written to the log. Grid legs - default 3, hard cap 5 Lot multiplier - fixed at 1.00x, cannot be raised Basket stop-loss - default 3.0% of equity, hard cap 10% Daily drawdown - default 4.0%, hard cap 20% Equity stop (latching) - default 15.0%, hard cap 50% Risk per trade - default 0.5%, hard cap 5% Margin usage - default 20%, hard cap 90% There is no martingale path through this code. Not a discouraged one - an absent one. The lot multiplier ceiling and floor are the same number, so every grid leg is the size of the first. The basket stop cannot be disabled. If you are evaluating a grid EA and the seller cannot tell you the ceiling, you are looking at an uncapped one. --- 2. A grid that refuses --- Most grid EAs add a leg because price moved. This one adds a leg only if the directional bias still agrees. When the bias has turned, the leg is refused and the reason is logged. Across the test set it added 201 legs and refused 228. Averaging into a reversal is how grid accounts die. --- 3. Eleven gates before any entry --- Spread ceiling. Median-spread multiple. Volatility-shock pause. Spike filter. Tick-density floor and ceiling. Chaotic-move block. Session windows. Day-of-week windows. Holiday flatten. News blackout (MT5 calendar or your own CSV). Daily trade cap. Every gate that blocks a trade says so, by name, in the log. You are never left wondering why the EA sat still. Most EAs fail silently; this one is auditable by construction. --- 4. Survives contact with reality --- Symbol verification on init: the EA identifies the instrument and, away from gold, warns in the log that every default was chosen for XAUUSD. Set AllowOtherSymbols to false if you want it to decline to trade there at all. Broker minimum-stop headroom on every order, so a one-tick widening does not become INVALID_STOPS. Order retries with prices refreshed on each attempt, not stale ones. Basket state persists across a terminal restart mid-position. Friday and holiday flatten. 139 inputs across 15 groups, every one range-clamped and cross-validated at startup, with the resolved configuration printed to the log. --- 5. The validation harness, and what it found --- The EA ships with a measurement method, and that method was turned on its own reference entries: 81.6 million real ticks (Sep 2025 - Sep 2026), split into two independent six-month blocks, every variant scored against a random-entry control. Ten entry families, five payoff ratios, four target sizes, six filter strengths, five total-cost levels. Result: none of the bundled entries beat chance. The largest t-statistic across the entire sweep was 1.54. Break-even sits at roughly 21 points of total cost, while a real raw gold account is 16-28 points all-in. That is published here, in the sales copy, on purpose. Why that is a reason to buy rather than not: any vendor can show you a curve. The question is whether they ran the test that could have embarrassed them. A random-entry control, two out-of-sample blocks and a t-statistic are exactly that test, and this is what it returned. You are buying the risk engine and the apparatus that produced that answer, including the ability to point it at your own entry and get an equally blunt verdict. One example of what the harness caught: MT5 custom symbols carry no commission in the tester. A raw account at 12 points spread plus 7 USD per lot round turn is a 19-point test, not a 12-point one. Measure it the naive way and you manufacture a profit that does not exist on any real account. --- Who this is for --- Developers with an entry idea who do not want to write a risk layer, a gate stack and a restart-safe basket manager from scratch. Traders running a grid or basket EA who want a hard ceiling they can actually verify. Anyone who has been burned by a backtest that did not survive live trading. --- Who this is NOT for --- If you want an EA that promises returns, buy something else. This makes no performance claim and never will. It does not generate signals. You supply the entry. It is not a martingale and cannot be configured into one. --- Requirements --- MetaTrader 5. XAUUSD, any suffix. Hedging or netting account. Any broker. No DLLs, no WebRequest, no external calls - nothing leaves your machine. Full documentation is included: what every module does and why, how to build a tick-level test bed, a testing checklist of around 70 cases, and the complete record of the testing above, failures included.